One of the first questions I am asked whenever I present my research usually is “what is neurofinance?”. Below is my view on this emerging field of research, defining it, especially with regards to behavioral finance and neuroeconomics.
The Neurofinance Paradigm
In neurofinance, we examine experimentally the nature of the cognitive processes engaged in acquiring and processing information in financial decision making. We further study how people select action plans based on the acquired representations of the values of potential investment prospects. One of our goals is to identify what kind of information the human brain can process efficiently (and what kind it cannot), as well as the environmental conditions facilitating or hampering this information processing. Another goal is to better understand how investment decisions are tuned depending on the appreciation of distinct kinds of uncertainty, such as risk, jump risk, and estimation uncertainty (ambiguity and model uncertainty).
A new Kind of Behavioral Finance
Behavioral finance emerged in the 90s to perfect the insights of mathematical finance. The point of departure of behavioral finance is that because classical finance assumes full rationality, it cannot explain many price patterns. Using insights from all behavioral sciences (cognitive neuroscience, psychology, sociology) on how real people depart from the rational model — real people are boundedly rational, behavioral finance can rationalize hitherto-puzzling price patterns.
The epistemology underlying neurofinance is different and reflects recent advances in decision neuroscience. We’re initially agnostic about the degree of rationality of people, i.e, we do not take people to be limited in their computational capabilities. Rather, we infer their degree of sophistication experimentally, from the observation of behavior and neural activity during cognitive tasks performed in the lab. These cognitive tasks replicate challenges that are routinely encountered in real world financial decision making. E.g., learning asset distributions that jump across time, or guessing what other people think—Theory of Mind.
One important aspect of this new paradigm is to examine in the lab which environmental conditions hamper the emergence of rationality, and which conditions help make people smart. Thus, Neurofinance affords a unique opportunity to
- develop acute prediction of investors’ behavior
- identify environmental markers of behavioral sophistication/irrationality in financial markets
- create nudges to aid decision making
Why looking at Neural Activity?
Methodology-wise, neurofinance lies at the intersection of experimental economics and computational neuroscience. We replicate in the lab core challenges faced by finance practitioners, and we examine how lab subjects (regular people as well as finance professionals) solve these challenges.
The question is: Do the cognitive processes that the subjects implement approximate the optimal solution, which “Mr Spock” (the rational agent) would implement? Or are these cognitive processes more akin to the boundedly rational heuristics which “Homer Simpson” would use1?
To answer this question, we do two things:
- Look at behavior: Sometimes, from observing the choices of a subject throughout the experiment, we can infer to what extent the subject acted more like Mr Spock, or more like Homer. This kind of inference works well when Mr Spock and Homer would behave differently in the task, which is often the case.
- Scan the brain of the subjects during the experiment: If we identify brain regions with a response profile consistent with the specific computational process performed by Mr Spock (resp Homer), the behavioral evidence that subjects acted more like Mr Spock (resp Homer) is strengthened.
One example: To learn optimally the expected returns of assets that jump over time, investors must acquire Bayesian jump detection signals, which they use at each point in time to tune their learning rate. The plausible alternative to this Bayesian learning, reinforcement learning, does nothing of the kind. So, identifying brain regions whose activity correlates with the Bayesian signals enables the inference that subjects approximated Bayesian learning. The inference is powerful because it is very unlikely that the identification of these neural signals be the result of serendipity.
Computational neuroeconomics Applied to Finance
This computational approach reflects a new trend in neuroeconomics. By identifying regions that implement a specific computational process, instead of merely reporting the “activation” of a brain region in a given experimental condition (which involves many computational processes), this approach enables a more convincing form of inference than is traditionally made in functional imaging studies.
What for? Implications for the industry
Portfolio managers and traders have to process information on the spot in rapidly changing environments. Little is known about how to tailor organizational and individual decision-making processes to help people process information efficiently in such contexts. By identifying environmental factors improving efficient information processing, it is hoped that research in neurofinance will produce practical results on how to improve investment and trading decisions, at both individual and organizational levels.
